Earlier quoted context omitted.
I think we need to acknowledge that a lot of what these agreements outline is, to a large degree, subjective. If human language, or expression in general, weren't contextual and up for interpretation, then I think this would be a viable way of looking at things, but that simply isn't the case. So long as there is room for interpretation, and the consequences (politically, economically, etc.) of those interpretations…
So you have n companies each applying their own subjective judgement over their own networks, or you have 1 government applying its own subjective judgement over all networks in that country. It's highly unlikely that the latter is more friendly to fringe or extremist views (setting aside whether or not that's a good thing).
This is empirically false though, at least in the the USA.
In the USA there are extremely strong protections for free speech.
If it was required that platforms followed the existing way that the government is allowed to ban certain speech, then platforms would be much less restrictive, because our laws of what speech is allowed is very pro free speech.